AI NewsInfrastructureAnnouncement
Google publishes a help page for letting Google-Agent reach your site
Google published a help document on how to let Google-Agent traffic through a CDN, firewall or bot-detection service.

Why it mattersSites that block bots will also block Google's shopping, booking and on-site search agent. Owners have to add it to the allow list.
A real buyer can send an AI agent to click Buy or Book on your site. The same firewall rule that keeps scrapers out can stop that agent. Google has now published a help document for that case. It tells site owners how to let Google-Agent through their content delivery network, their firewall and their bot-detection service.
Google defines Google-Agent as the name its agents use when they navigate the web and perform actions for a user. The agents work on Google infrastructure. Barry Schwartz at Search Engine Roundtable reported the help page as a new one from Google.
What the page tells you to do
Google says to allow Google-Agent in the layers that see traffic first. Those are the CDN, the web application firewall and the bot-detection service. The help page lists three ways to confirm the request is from Google. The first is the User-Agent header. Every request contains Google-Agent in its value. The second is the IP address. Google publishes the current list of IP ranges at developers.google.com/static/crawling/ipranges/user-triggered-agents.json. Every verified request comes from one of those addresses.
Signed requests
The third way is a signature. Google is testing the Web Bot Auth protocol. That is a standard called RFC 9421 for signing HTTP messages. The signing identity is https://agent.bot.goog. Sites can read Google's public keys at https://agent.bot.goog/.well-known/http-message-signatures-directory. A site that trusts the signature can let the request past bot checks on that proof alone.
What counts as an agent action
Google lists the actions its agent performs for a user. Those are on-site search, filtering, and clicking into detail pages. The list also covers add-to-cart and booking. The help page says that bot-detection services, CDNs and firewalls already provide ways to allow this traffic. Site owners have to turn the allow rule on in their own provider. Google tells them to refer to the provider for the steps.
The open web is adding a second class of traffic. An AI agent that acts for one real buyer is now one of its visitors. A site built for human visitors only will miss those sessions. Owners who sell online should treat Google-Agent as a real user. Turn the block off before a buyer notices nothing works. Then test an order and a booking with the agent, and look for the request in the gateway log. The signed request path is the one worth watching. It needs no IP list refresh when Google adds a new range.
Source
Google-Agent help document from Google. New Google Search Help Document On How To Allow Google-Agent Traffic by Barry Schwartz at Search Engine Roundtable.
This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.
Get AI News in your inbox
New developer tools, model and agent releases, and how teams are actually using them to release software. Short, and only when there is something worth reading.


